Victor de Mello Lecture - Soils and RocksSoils and Rocks v. 37, n. 2 Victor de Mello Lecture The Victor de Mello Lecture was established in 2008 by the Brazilian Associa- tion for

Soils and Rocksv. 37, n. 2

Victor de Mello Lecture

The Victor de Mello Lecture was established in 2008 by the Brazilian Associa-tion for Soil Mechanics and Geotechnical Engineering (ABMS), the BrazilianAssociation for Engineering Geology and the Environment (ABGE) and the Por-tuguese Geotechnical Society (SPG) to celebrate the life and professional contri-butions of Prof. Victor de Mello. Prof. de Mello was a consultant and academic forover 5 decades and made important contributions to the advance of geotechnicalengineering. Every second year a worldwide acknowledged geotechnical expert isinvited to deliver this special lecture, on occasion of the main conferences ofABMS and SPG. The fourth Victor de Mello Lecture is delivered by Prof. JamesK. Mitchell, whose name requires no introduction as author, professor, researcherand engineer. Prof. de Mello used to exchange ideas with his friend, Prof. Mitchell,most recently on topics related to the book he was writing. Prof. Mitchell chose forthe lecture a theme that was one of Prof. de Mello's passions: dams. In his thoroughreview of the upgrading process of two old dams, focused on effective risk mitiga-tion, Prof. Mitchell revisits and expands many of the concepts put forth by Prof. deMello in his Rankine Lecture of 1977.

Prof. JAMES K. MITCHELL is University Distinguished Professor, Emeritus,at Virginia Tech and Consulting Geotechnical Engineer. He received his B.C.E.from R.P.I. in 1951, and M.S. and Sc.D. degrees from M.I.T. in 1953 and 1956. Hewas a professor at the University of California, Berkeley, from 1958 to 1994 be-fore moving to Virginia Tech, and was Chairman of the Civil Engineering Depart-ment there during 1979-1984. His teaching, research and consulting have focusedon soil properties and behavior, soil stabilization and improvement, ground rein-forcement, in-situ measurement of soil properties, and mitigation of ground fail-ure risk during earthquakes. He has authored 400 publications, including threeeditions of the book, "Fundamentals of Soil Behavior." Dr. Mitchell is a Distin-guished Member of the American Society of Civil Engineers and was Vice Presi-dent of the International Society for Soil Mechanics and GeotechnicalEngineering from 1989-1994. He is a member of the United States National Acad-emies of Engineering and Sciences.

Lessons From The Lives of Two Dams

J.K. Mitchell

Abstract. Many embankment dams completed during the first six decades of the 20th century have been found deficientrelative their ability to resist currently anticipated levels of seismic shaking and probable maximum flood. In this FourthVictor de Mello Lecture, two recent case histories are described. One is a hydraulic fill structure completed in 1920 that isfounded on alluvial material, some zones of which are susceptible to liquefaction. The other is a zoned earthfill damcompleted in 1956 that is founded over a channel filled with loose, uncompacted, hydraulically placed tailings from goldmining operations. Each dam has been upgraded in phases over periods of several decades using different strategies andground improvement technologies to improve stability and reduce failure risks. Several take away lessons from theseexperiences concerning current risk mitigation strategies, the importance of correct soil and site characterization, andimplementation and effectiveness of different ground stabilization and improvement methods are presented.Keywords: earthquakes, embankment dams, liquefaction, risk mitigation, site characterization, soil improvement.

1. Introduction

Dr. Victor de Mello was one of the “Giants of Geo-technics” of the 20th century who leaves a legacy of excep-tional contributions from his major accomplishments as apracticing civil and geotechnical engineer, as a teacher bothinside and outside the classroom, as a researcher, as a leaderin his profession, and as a dynamic, yet philosophical andcongenial colleague and friend. I am greatly honored by theinvitation to deliver this lecture in celebration of his life andprofessional contributions, while at the same time dauntedand humbled by the challenge of contributing somethingworthy of the honor.

Victor de Mello devoted his 1977 Rankine Lecture toconsiderations in embankment dam design (de Mello,1977), with special focus on filters, drainage and seepagecontrol, as well as stability issues. He discussed factors ofsafety and their meaning, introducing considerations ofprobability and variability that, while relatively new then,are now central to assessment of dam safety. I have alsochosen to address embankment dams, but in this lecture, thefocus is on dealing with problems that arise in existingdams resulting from age and from risks caused by extremeevents, especially earthquakes and floods, that were incom-pletely understood and accounted for at the time of originalconstruction many decades ago.

Many large embankment dams were constructed inthe U.S.A. during the first six decades of the 20th Century.The M 9.2 Great Alaska Earthquake and the M 7.5 NiigataEarthquake in Japan, both in 1964, focused attention on soilliquefaction, and the near catastrophic failure of the LowerSan Fernando Dam in southern California in the 1971 M6.6 San Fernando Earthquake led to reevaluation of theseismic vulnerability of many other dams. The MaximumCredible Earthquakes, Maximum Probable Floods, andpopulations at risk have increased significantly at many

sites. Risk analyses have led to unacceptably high potentialconsequences requiring implementation of mitigation mea-sures at many dams. Two of these dams are described inthis paper.

San Pablo Dam, near Oakland, California and com-pleted in 1921, is a hydraulic fill structure founded on allu-vial deposits. Mormon Island Auxiliary Dam (MIAD), nearSacramento, California, is a compacted fill embankmentfounded on hydraulically deposited dredger tailings result-ing from gold mining operations; completed in 1956. Eachdam was subsequently deemed unsafe under the anticipatedseismic loading conditions. Several modifications havebeen made to each dam to improve resistance to anticipatedearthquake loadings and updated flood risks. These modifi-cations took place from 1967 to 2010 at San Pablo Dam andhave extended from the late 1980’s to a planned final com-pletion in 2016 at the Mormon Island Auxiliary Dam. Someconclusions and lessons learned about the development ofgeotechnical earthquake engineering for dams, seismicremediation strategies, the importance of proper site andmaterial characterization, and the advantages and limita-tions of some ground improvement methods can be derivedfrom these two case histories.

2. San Pablo Dam

This dam is a 53.3 m (170 ft) high, 38.1 m (125 ft)crest width, 366 m (1200 ft) long hydraulic fill damfounded on alluvial sediments that contain some zones thatare susceptible to liquefaction. The hydraulic fill materialused for construction of the embankment consists of weath-ered sandstone and shale that was obtained from the EastBay Hills near Oakland, California. The site is locatedwithin a few kilometers of several major faults, and it is es-timated that there is a 62 percent probability of one or moreearthquakes of magnitude 6.7 or greater during the period2003 to 2032.

Soils and Rocks, São Paulo, 37(2): 99-109, May-August, 2014. 99

J.K. Mitchell, Sc.D., P.E. University Distinguished Professor, Emeritus, Virginia Tech, Blacksburg, VA 24061-0105, U.S.A. e-mail: [email protected].

Photos of the construction of the dam illustrating theexcavation for the core trench and placement of the hydrau-lic fill embankment materials are shown in Fig. 1. The orig-inal hydraulic fill embankment construction was completedin 1921. Tests on a few sandy samples of the embankmentmaterials in the 1960’s and 1970’s indicated potentiallyliquefiable behavior. Evidently this led to the assumption ofa liquefiable embankment because it was a hydraulic fill,and hydraulic fills of cohesionless materials are invariablyof low relative density and high liquefaction potential un-less densified following deposition.

A small downstream buttress fill was constructed in1967 to improve seismic stability. Then, following the1971 San Fernando Earthquake in southern California, amuch larger upstream compacted fill buttress extending tobedrock was completed in 1979. Construction of this but-tress required draining the reservoir, with its attendant de-pletion of the area’s water supply capacity and disruption ofrecreational use of the reservoir. A cross section of the max-imum embankment section with both the 1967 downstreamstabilizing berm and the much larger 1979 upstream bermin place is shown in Fig. 2.

A new seismic stability evaluation was completed in2004 assuming a liquefiable embankment and a M7.25earthquake on the Hayward Fault which passes just 3 kmsouthwest of the dam. The results indicated the potential forvertical slumping of up to 10.7 m and overtopping of thedam by the impounded reservoir. To provide additionalfreeboard for the short term, the reservoir level was loweredby 6 m. Consideration was given to completely rebuildingthe dam; however, this would have required again drainingthe reservoir, an action that had considerable opposition.Instead, an in-place alternative was chosen that consisted ofCement Deep Soil Mixing (CDSM) to bedrock at depths ofup to 36.5 m in the downstream foundation alluvial soilsand construction of a large buttress fill on the downstreamembankment slope, shown as Initial Remedial Concept inFig. 3, from Yiadom & Roussel (2012).

An extensive new field investigation program wasthen completed that included many cone penetration tests(CPT) and borings into both the embankment and founda-tion materials. The results of these field investigations andlaboratory tests on representative samples are described indetail by Moriwaki, et al. (2008). They described the sam-pled hydraulically placed material as consistently veryclayey, very “lumpy”, and over-consolidated. Of special in-terest is Fig. 4, taken from that reference, which shows CPTand plasticity data for the hydraulic fill embankment shellmaterials, zones (1) and (2) in Fig. 2. This data showsclearly that this material is not susceptible to liquefaction,as had been assumed for the previous evaluations of SanPablo Dam and for the Initial Remedial Concept in Fig. 3.In retrospect this finding is not surprising given that the ma-terial used for the hydraulic fill came from the colluvial

slopes of the surrounding hills where the soil type is knownto be largely silty and clayey.

This re-characterization of the hydraulic fill embank-ment material from liquefiable to non-liquefiable, fine-grained soil enabled significant reductions in the requiredsizes of the 2010 buttress and CDSM block, as may be seenin Fig. 3 by comparing the Final Remediation Design withthe Initial Remedial Concept. Seismic deformation analy-ses of the maximum composite dam cross-section weredone using computer program FLAC with input groundmotions that had a peak acceleration of 0.98 g and associ-

100 Soils and Rocks, São Paulo, 37(2): 99-109, May-August, 2014.

Mitchell

Figure 1 - Stages in the construction of San Pablo Dam (East BayMunicipal District construction photos Reproduced and reportedby TNM Terra Engineers, Inc., Ninyo and Moore, 2007). (a) Coretrench to bedrock with shoring in 1917; (b) Hydraulic fill con-struction site in 1919; (c) Hydraulic fill transportation and deposi-tion in 1920.

ated spectral content consistent with the occurrence of a M7earthquake on the nearby Hayward Fault. The results indi-cated maximum seismically induced permanent displace-ments in the range of 0.3 to 0.6 m (1 to 2 ft) (Kirby et al.,2010), which were considered well within the range for ac-ceptable performance.

The reduced volumes of required CDSM and the newdownstream buttress fill realized a cost saving of about US$40 million, a very significant amount for a project with aconstruction cost of about $60 million. Kirby et al. (2010)describe the design of the CDSM foundation block, theconstruction process, and the methods used for quality con-



Figure 2 - Composite maximum section of San Pablo dam after buttress additions. (adapted from Moriwaki et al., 2008). (1) upstreamshell (hydraulic fill); (2) downstream shell (hydraulic fill); (3) ponded clay/silt; (4) core and key trench (hydraulic fill); (5) alluvial foun-dation soils; (6) upstream buttress (well-compacted) completed in 1979; (7) downstream buttress (less-compacted) added in 1967; and(8) bedrock.

Figure 3 - Remedial designs for mitigation of seismic risk to San Pablo Dam (from Yiadom & Roussel, 2012).

Figure 4 - CPT and Classification data for the San Pablo Dam hydraulic fill shell material. Zone A: Cyclic liquefaction possible; Zone B:Cyclic liquefaction unlikely; Zone C: Flow/cyclic liquefaction possible (from Moriwaki et al., 2008).

trol of the material. Another cost saving feature was that thespoils from the deep cement mixing could be incorporatedinto the new downstream buttress. An aerial photo (Google,2012) of the project after completion is shown in Fig. 5.This project is an excellent illustration of the importance ofcorrect soil identification and classification prior to analy-sis and design of risk mitigation strategies for existingdams.

3. Mormon Island Auxiliary DamThe Mormon Island Auxiliary Dam (MIAD) forms a

part of the Folsom Project located on the American Riverabout 32 km (20 miles) northeast of Sacramento, Califor-nia. This project provides water supply, hydroelectricpower, and flood protection for a large metropolitan area. Itconsists of a concrete main dam, right and left wing dams,the zoned and rolled earthfill MIAD, and eight earthfilldikes that are needed to contain Folsom Lake. A plan show-ing these features is given in Fig. 6. Figure 7 is an aerialphoto showing MIAD as it appears at present (GoogleEarth, 2013).

MIAD was constructed by the U.S. Army Corps ofEngineers and completed in 1953, after which operationand maintenance activities have been the responsibility ofthe U.S. Bureau of Reclamation. The dam is comprised of athin central impervious core bounded by fine and coarse fil-ter transition zones extending to weathered metamorphicbedrock and compacted earthfill shells both upstream anddownstream, as shown in Fig. 8. The embankment shellsare composed of alluvium dredged from the site. The de-sign width of the dam crest is 7.77 m (25.5 ft), the upstreamslope varies from 3H:1V to 4.5H:1V, the downstream slopevaries from 2.5H:1V to 3.5H:1V. The structural height ofthe embankment is 50.3 m (165 ft), and the dam is 1470 m(4820 ft) long.

The alluvial foundation materials, consisting of vary-ing amounts of gravels, sands, silts and clays, were dredgedand re-dredged to depths of very near the bedrock along anapproximately 300 m (900 ft) long strip adjacent to the

present location of the left abutment of the dam as a part ofgold mining operations during the latter part of the 19thCentury. As a result of this dredging and re-deposition,these materials were left in a loose state. The upstream anddownstream toes of the embankment are underlain by anapproximately 20 m (60 ft) thick layer of loose tailings for


Mitchell

Figure 5 - San Pablo Dam in 2010 after seismic remediation usingcement deep soil mix in the downstream foundation and a down-stream stability berm (Google Earth, Imagery Date 10/2/2009).

Figure 6 - Components of the Folsom Project near Sacramento,California.

Figure 7 - Arial view of Mormon Island Auxiliary Dam as it ap-peared in August 2013. Critical section over potentiallyliquefiable foundation material extends about 300 m from the leftabutment (Google Earth, Imagery Date 8/13/2013).

Figure 8 - Cross section of the Mormon Island Auxiliary Dam asconstructed in 1956.

this 300 m length of the dam. Evaluation of the seismicsafety of the dam during the 1980s indicated that liquefac-tion of these tailings was likely during the design earth-quake. A series of dam and foundation modifications forrisk mitigation have been undertaken since then, and theseactivities are described below.

The results of field and laboratory tests were used todevelop the distributions of normalized SPT (StandardPenetration Test) values of (N1)60 in the foundation that areshown in Fig. 9. These values were used to evaluate factorsof safety against liquefaction and for the establishment ofparameters needed for dynamic deformation analyses.Ground improvement was implemented beneath both theupstream and downstream embankment toe areas to miti-gate the liquefaction risk and to limit dynamic displace-ments.

Owing to severe drought conditions in California inthe late 1980s the reservoir level was low, and this made itpossible to undertake deep dynamic compaction (DDC) inthe dry from the upstream embankment. As shown in Fig.10, an access excavation was made to provide level groundfor carrying out the work. A block of densified soil wasformed by repeated dropping of a 2.0 m (6.5 ft) diametersteel drop weight of 31.75 tonnes (35 tons) over a 244 by46 m (800 by 150 ft) treatment area. The drop height of32.9 m (108 ft) corresponded to a free fall distance of 30 m(98.4 ft). Three coverages of the area were made, with30 drops at 15.2 m (50 ft) center to center drop point spac-ing for the first, or primary, coverage, 30 drops at pointssplitting the primary coverage spacing for the second cov-erage, and 15 drops at points splitting the secondary cover-age for the third coverage. Finally surface “ironing” wasaccomplished using 2 weight drops from a height of 10 m(30 ft) at adjacent points to cover the entire area. Finally,the access excavation was re-filled, and a post treatmentberm was placed on the upstream slope, as shown schemati-cally in Fig. 10. The photograph in Fig. 7 shows this bermextending from the upstream slope into the reservoir fol-lowing subsequent increase in the reservoir water level.

The effectiveness of the DDC can be seen in Fig. 11,which shows values of Becker Penetration Test (BPT) re-

sistance, converted to equivalent SPT values of (N1)60, as afunction of elevation. The BPT measures the number ofblows to drive a 168 mm (6.6 in) outside diameter dou-ble-walled casing a distance of 0.3 m using a double-actingdiesel pile driving hammer. This test is useful in soils con-taining gravel and cobbles, where difficulties are often en-countered when using the SPT. The decrease in penetrationresistance with depth in Fig. 11 is characteristic of the DDCmethod for soil improvement, with an effective treatmentdepth of about 10 m (35 ft) being about the maximum at-tainable when heavy weights are used in soils of no to lowplasticity. As shown in Fig. 11, the DDC was unable todensify the soil within the treatment zone over the fulldepth to the underlying bedrock.

Following an extensive testing program for determi-nation of the most suitable methods for in-place improve-ment of the downstream foundation material, the systemshown schematically in Fig. 12 was designed, with con-struction completed during 1993-1994. The improvementzone was a strip along the downstream toe of the dam that is



Figure 9 - Characterization of the MIAD foundation material interms of SPT values of (N1)60 for use in liquefaction potential as-sessments.

Figure 10 - Ground improvement design intended for mitigationof liquefaction risk beneath upstream embankment of Mormon Is-land Auxiliary Dam. DDC completed in 1990.

Figure 11 - Pre- and post-deep dynamic compaction Becker Pene-tration Test equivalent (N1)60 values beneath the upstream slope ofMormon Island Auxiliary Dam.

900 ft (275 m) long by 200 ft (61 m) wide in plan, in the arealabeled “Downstream improvement” in Fig. 7. Excavationinto the existing downstream embankment was required inorder to develop a level working platform, Fig. 13, for in-stallation of 1.2 m (4 ft) diameter wet, bottom feed, vibro-replacement stone columns and the upstream and down-stream drainage zones that are composed of 250 mm diam-eter gravel columns on 1.0 m centers installed using avibro-pipe. It is important to note that excavation into thedownstream slopes of embankment dams is not withoutrisk owing to reduced seepage paths and lower factors ofsafety against stability failure while the excavation is open,both of which must be accounted for in assessing risk dur-ing construction.

The primary purpose of the stone columns was todensify the loose liquefiable dredged alluvium foundationmaterial so that it would not liquefy under the design earth-quake. The upstream and downstream drainage zones areintended to intercept pore pressure plumes migrating to-wards the stone column treated zone in the event liquefac-tion develops in the adjacent untreated dredged alluviumduring an earthquake. Prevention of pore pressure increasesin the stone column zone is important for maintenance ofshear strength during and after shaking.

The profile through the dredge spoils in the stone col-umn area indicated that the upper 4.5 to 6 m contains coarsesand to cobble size material, and the lower 3 to 6 m is siltysand to silty clay with 10 to 77 percent fines, with an aver-age fines content of 30 percent (Allen et al., 1995). The pre-and post-treatment penetration resistance values shown inFig. 14 are consistent with these conditions. A cross-section of the embankment showing all the modificationsand ground improvement work through 1994 is shown inFig. 15.

Subsequent investigations and risk analyses were ini-tiated in 2001. These studies took into account that a poten-

tially liquefiable zone remains beneath the upstream blockof material that had been densified by deep dynamic com-paction. After extensive re-analyses of available data andfurther in-situ testing, it was concluded also that the neces-sary downstream foundation densification was notachieved in the lower part of the soil profile during installa-tion of the downstream stone columns, owing primarily tothe fine-grained nature of the material. Furthermore, con-tamination of the stone columns by the fines resulted inlower column strength than had been anticipated and im-peded the drainage capability as well. These re-analyses ledto the conclusion that the most critical seismic failure mode


Mitchell

Figure 12 - Downstream foundation improvement for reduction of seismic failure risk at Mormon Island Auxiliary Dam. Constructionwas completed in 1994.

Figure 13 - Installation of stone columns and gravel mini-columndrains beneath the downstream embankment of Mormon IslandAuxiliary Dam in 1994. Note the excavation and steepened slopeneeded for development of a working platform and for optimiza-tion of the improvement zone location for its effectiveness as akey block.

would result from liquefaction of the upstream and down-stream foundation materials leading to significant deforma-tions of the dam in the downstream direction, with verticaldisplacements sufficient to result in overtopping for highreservoir levels.

A risk analysis done in 2007 showed that both the An-nual Failure Probability and the Annualized Life Loss wereabove the Bureau of Reclamation guideline values. Correc-tive Action Studies for Seismic and Static Risk Reductioncompleted in 2010 led to a design that included a concretekey block with compacted soil above in the downstream

stone column area and a compacted soil buttress fill overthe downstream embankment slope that includes underly-ing filters and drains. No further remedial work is plannedfor mitigation of the upstream liquefaction risk; i.e., the keyblock and buttress design is considered sufficiently robustthat if any deformations develop in the upstream direction,the remaining downstream core, filters, and berm will stillbe adequate to maintain stability and the necessary free-board.

An additional point of interest is that jet grouting wasproposed initially for construction of the key block. How-ever, the results of an extensive test program indicated thatthe required continuity and strength of soilcrete could notbe obtained owing to the presence of gravel, cobbles, andstone columns in the treatment zone. Jet grouting was ulti-mately judged to be technically and economically unfeasi-ble for this purpose at this site. As a result, construction ofthe key block was accomplished within a series of contigu-ous cells, each being formed within an open, braced exca-vation. A photograph of this work in progress is shown inFig. 16.

The final configuration of MIAD will be as shownschematically in Fig. 17. The key block construction wascompleted in February 2013, and the buttress fill is sched-uled for completion in 2016.

4. Take-Away Lessons From These TwoProjects

Several lessons and conclusions can be drawn fromthe San Pablo Dam and Mormon Island Auxiliary Damseismic remediation work, and other projects within the au-



Figure 14 - Penetration resistance profiles in the downstreamstone column treatment area, Mormon Island Auxiliary Dam.

Figure 15 - Cross-section of Mormon Island Auxiliary Dam showing conditions after modifications completed in 1994 (adapted fromU.S. Bureau of Reclamation, 2010).

thor’s experience, that relate to geotechnical engineering ofexisting embankment dams, strategies for mitigation ofrisk, site and material characterization, and ground im-provement methods and their applicability.

4.1. Geotechnical engineering and failure risk mitiga-tion for existing embankment dams

Most existing large embankment dams in the U.S.were constructed prior to the 1960s. Little attention waspaid to seismic issues in their design. However, large earth-quakes in Alaska and Japan in 1964, and the failure of theLower San Fernando Dam in California in 1971 triggeredassessment of the seismic resistance of major dams. Manydams have required modifications to protect against crack-ing and excessive deformations in the event of future earth-quakes. In addition, seismicity reevaluations and redefi-nitions of maximum credible earthquake at a site, alongwith increases in the probable maximum flood and the pop-ulations at risk have resulted in increased demands.

Potential failure mode analyses and formal risk analy-ses are now widely used for determining the urgency of un-

dertaking risk mitigation activities, evaluation of the effec-tiveness of various types of remediation to be employed,and for prioritizing projects within available time and bud-get. Risk management guidelines and details of the risk as-sessment and evaluation process have been developed byseveral water resources and regulatory agencies. Publica-tions by the U.S. Bureau of Reclamation and by the U.S.Army Corps of Engineers, many of which are availableon-line, provide extensive information on these proceduresand interpretation of the results.

A reasonable goal is to bring the safety of the dam, asmeasured by global stability, resistance to deformation suf-ficient to prevent overtopping resulting from excessivecrest settlement, filter protection, safety against cracking,and drainage provisions downstream of the seepage barrierto states that are as safe as would be attainable if the damwere being designed and constructed today.

Remediation strategies for achieving the needed lev-els of stability, mitigation of liquefaction. controllingdeformations under seismic loading, and prevention ofovertopping seem to have followed a path over the past 50


Mitchell

Figure 16 - Schematic diagram and photos of the excavation and bracing system for construction of the concrete key block in the down-stream toe area of Mormon Island Auxiliary Dam (adapted from U.S. Bureau of Reclamation).

years or so from adding a simple buttress fill, to incorporat-ing different types of in-situ ground improvement such asdeep dynamic compaction, vibro-compaction, vibro-repla-cement, and/or compaction piles both upstream and down-stream, to a focus on downstream work only.

The downstream only option avoids the need forworking over and through water, unless the reservoir levelcan be drawn down. An upstream embankment failure canbe allowed provided the downstream embankment is but-tressed sufficiently to prevent excessive loss of freeboardand the upstream failure zone does not encroach on the damcore at a point beneath the reservoir level. Satisfying theseconditions must be demonstrated by suitable analyses. Adownstream buttress fill overlay above a block of founda-tion soil treated in-situ, as done for the San Pablo Dam, issimple and reliable. The same is true for the buttress fillover a key block, now under construction at the Mormon Is-land Auxiliary Dam.

It is imperative to keep in mind, however, that when-ever a project involves excavation into the downstreamslope; e.g., during the stone column and drain installation,Fig. 13, or into the foundation, as was necessary for the keyblock construction at MIAD, Fig.16, there may be in-creased seepage as well as a temporary decrease in factor ofsafety against stability failure while the excavation is openthat must be evaluated and assessed in terms of increasedfailure risk.

Limit equilibrium and deformation analyses underboth static and dynamic loading conditions are necessary.These analyses, if properly carried out, provide critical in-formation about the current condition of the embankment,the potential impacts of different seismic or other new load-ing conditions, the locations of potential failure surfacesand large deformation zones and patterns, whether the de-formations may have detrimental effects on the dam core,filters, and seepage control components, and the effective-ness of different remediation methods and designs in assur-ing that adequate stability and deformation limits can beachieved.

A number of readily available computer programs isavailable for the limit equilibrium evaluations; the defor-mation analyses are usually done using finite difference;

e.g., FLAC and/or finite element; e.g. PLAXIS programs.The most critical input parameter for any of these analysesis the shear strength (Duncan, 2013). Determination of theappropriate strength value is often challenging, especiallyin situations involving liquefaction where knowledge of thepost-triggering strength of the liquefied soil is required inorder to assess the potential consequences.

4.2. Site and material characterization

The validity and reliability of all the analyses, selec-tion of risk mitigation methods, and predictions of futurebehavior hinge on proper knowledge of the subsurface ma-terials, their boundaries, the groundwater conditions, andhow the relevant properties are measured and assigned. Re-view of available, original geological and geotechnical re-ports and construction records is essential. Informationabout past modifications to the dam must be carefully as-sessed. At the same time, the information about the actualpresent characteristics of the embankment and foundationsoils may not be available or totally correct, as was found tobe the case at San Pablo Dam.

Incorrect identification and characterization of mate-rials can lead to significant overestimates or underestimatesof both the dam safety and the needed extent, time, and costof ground improvement.

4.3. Ground improvement methods and their applica-tions in embankment dams

Ground improvement is now a major sub-disciplinewithin geotechnical engineering and geo-construction.There are many methods and materials that can be used tomeet a variety of ground improvement and reinforcementapplications in embankment dams. A comprehensive de-scription and classification of methods was developed bythe ISSMGE Technical Committee on Ground Improve-ment (Chu et al., 2009). An open access web-site,GeoTechTools.org, described by Schaefer et al. (2012) isnow available that provides information and interactive se-lection guidance on 46 technologies useful for soil stabili-zation, ground improvement and reinforcement, and geo-construction. The information provided for each technol-ogy includes a technology fact sheet, photos, case histories,



Figure 17 - Mormon Island Auxiliary Dam after completion (est. 2016) of modifications to assure seismic safety (adapted from U.S. Bu-reau of Reclamation).

design guidance, quality control and quality assurance in-formation, cost information, specifications, and a bibliog-raphy.

The choice of the most appropriate ground improve-ment method or methods for mitigation of failure risks todams is critical. In retrospect, the use of deep dynamic com-paction for upstream foundation improvement at MIADwas later deemed not effective because of the unimprovedzone that remained beneath the densified block, Fig. 10 andZone 8 in Fig. 15. The presence of the high fines content inthe lower portion of the downstream foundation soil atMIAD, Fig. 14 and Zone 12 in Fig. 15, was later deemed re-sponsible for the stone columns to provide the improve-ment needed.

A few general observations concerning trends in theuse of different ground improvement methods for mitiga-tion of liquefaction risk and excessive deformations in damfoundations are:• The use of vibro-compaction and vibro-replacement is

decreasing• The use of deep soil mixing is increasing.• Buttress fills and downstream overlay fills are among the

most cost-effective treatment methods provided suitablefill material and the necessary space are available.

• The promise of jet grouting for use in dam foundations isyet to be realized.

• What you can see, measure, and test is invariably a betterand more reliable option than what you can’t see, pro-vided cost and construction risks are acceptable.

5. Some Continuing and UnresolvedProblems

A number of unknowns, uncertainties, and problemscan be identified that, if resolved, could lead to better riskevaluations, more optimized selection of mitigation strate-gies and ground treatment methods, and improved predic-tions of future behavior. A desirable goal is to “get it rightthe first time” so that subsequent mitigation measures, aswere needed at both San Pablo Dam and Mormon IslandAuxiliary Dam, will not be necessary. Among these un-knowns, uncertainties and problems are:• Anticipating (predicting) future increases in demand on

the facility; e.g., greater seismic loading, larger probablemaximum flood, adverse consequences of climate chan-ge, increased population at risk

• Interpreting and communicating the results of a riskanalysis

• Deciding the acceptable level of risk• Assessing the liquefaction potential of soils containing

gravel and cobbles• Assessing the liquefaction potential of silty soils• Assessing the post-earthquake residual strength of lique-

fied soil

• Selecting and implementing the appropriate soil consti-tutive model for use in liquefaction and dynamic defor-mation analyses

• Assessing the reliability and accuracy of dynamic defor-mation analyses. A widely accepted “rule of thumb” hasbeen that actual deformations may be within about +/-100 percent of the computed or estimated value; how-ever, it is not really known how valid this estimate is.

• Accounting for time (aging) effects following densi-fication and/or admixture stabilization of foundation andembankment soils

• Writing enforceable specifications that will produce theneeded end results, but also allow for inherent variabilityin materials and other site-specific conditions

• Assessing compliance with the ground improvementspecifications for uniformity and post-treatment strengthand stiffness requirements

6. Concluding Comments

Getting it right the first time can be very difficultgiven the unknowns and uncertainties at the time of initialdesign and construction of an embankment dam. The twocase histories described in this paper illustrate that gettingthe remediation of a deficient existing dam right the firsttime can also be very difficult. The potential consequencesof climate change, increasing numbers and magnitudes ofextreme events (floods, storms, earthquakes, fires, etc.)must be considered from the outset of a project. Simple, ob-servable, and measurable methods for dam strengtheningand risk mitigation should be used wherever possible. Areasonable overall goal should be to make an existing, defi-cient embankment dam as safe as if you were starting a newproject today. Resolution of the issues listed in the previoussection should be instrumental in helping to reach this goalby enabling better selection and optimization of methodsfor mitigation of risks to existing dams in the future.

References

Allen, M.G.; Jones, R. & Gularte, F.B. (1995) Bottom-feedstone columns, wet replacement construction method:Mormon Island Auxiliary Dam modifications. Soil Im-provement for Earthquake Hazard Mitigation,Geotechnical Special Publication 49, ASCE, p. 82-95.

Chu, S.; Varaksin, S.; Klotz, U., & Menge, P. (2009) Con-struction processes. Proc. Int. Conf. on Soil Mechanicsand Geotechnical Engineering, Alexandria, Egypt, IOSPress.

de Mello, V.F.B. (1977) Reflections on design decisions ofpractical significance to embankment dams. The Seven-teenth Rankine Lecture, Geotechnique, v. 27:3, p.281-355.

Duncan, J.M. (2013) Slope stability then and now. Pro-ceedings of the GeoCongress 2013: Stability and Per-formance of Slopes and Embankments III, p. 2191-210.


Mitchell

Kirby, R.C.; Roussel, G.L.; Barneich, J.A.; Yiadom, A.B.& Todaro, S.M. (2010) Design and construction of seis-mic upgrades at San Pablo Dam using CDSM. Proc.U.S. Society of Dams Conference, p. 137-152.

Mitchell, J.K. (2008) Recent developments in ground im-provement for mitigation of seismic risk to existing em-bankment dams. Geotechnical Earthquake Engineeringand Soil Dynamics IV, ASCE Geotechnical SpecialPublication 181, p. 1-20.

Moriwaki, Y.; Dinsick, A.; Barneich, J.; Roussel, G.;Yiadom, A.; Starr, F., & Tan, P. (2008) Seismic charac-terization and its limited implication for San PabloDam. Geotechnical Earthquake Engineering and SoilDynamics IV, p. 1-10.

Schaefer, V.R.; Mitchell, J.K.; Berg, R.R.; Filz, G.M &Douglas, S.C. (2012) Ground improvement in the 21stcentury: a comprehensive web-based information sys-tem. Rollins, K. & Zekkos, D. (eds) Geotechnical Engi-neering State of the Art and Practice. ASCEGeotechnical Special Pub. No. 226, p. 272-293.

TNM Terra Engineers Ninyo and Moore (2007) Site char-acterization of Dam Area, Report, Design Services forSan Pablo Dam Seismic Upgrade, April 2007, 347 p.

U.S. Bureau of Reclamation (2010) Mormon Island Auxil-iary Dam Modification Project, Final Supplemental En-vironmental Impact Statement/Environmental ImpactReport. Folsom California, Mid-Pacific Region, StateClearing House #2009042077. May 2010.

Yiadom. A.B. & Roussell, G.L. (2012) Seismic upgrades ofSan Pablo Dam. Geo-Strata, May/June 2012, p. 34-39.



Documents

Victor de Mello Lecture - Soils and RocksSoils and Rocks v. 37, n. 2 Victor de Mello Lecture The Victor de Mello Lecture was established in 2008 by the Brazilian Associa- tion for